The vast majority of a bacterial population is killed when treated with a lethal concentration of antibiotics. The time scale of this killing is often comparable with the bacterial generation time before the addition of antibiotics. Yet, a small subpopulation typically survives for an extended period. However, the long-term killing dynamics of bacterial cells has not been fully quantified even in well-controlled laboratory conditions. We constructed a week-long killing assay and followed the survival fraction of Escherichia coli K12 exposed to a high concentration of ciprofloxacin. We found that long-term survivors were formed during exponential growth, with some cells surviving at least 7 d. The long-term dynamics contained at least three time scales, which greatly enhances predictions of the population survival time compared with the biphasic extrapolation from the short-term behavior. Furthermore, we observed a long memory effect of a brief starvation pulse, which was dependent on the (p)ppGpp synthase relA. Specifically, 1 h of carbon starvation before antibiotics exposure increased the surviving fraction by nearly 100-fold even after 4 d of ciprofloxacin treatment.